Support VITS models from icefall. (#625)

k2-fsa · Mar 1, 2024 · d569643 · d569643
1 parent 93836ff
commit d569643
Show file tree

Hide file tree

Showing 5 changed files with 13 additions and 4 deletions.
diff --git a/sherpa-onnx/csrc/offline-tts-vits-impl.h b/sherpa-onnx/csrc/offline-tts-vits-impl.h
@@ -205,7 +205,8 @@ class OfflineTtsVitsImpl : public OfflineTtsImpl {
     if (meta_data.frontend == "characters") {
       frontend_ = std::make_unique<OfflineTtsCharacterFrontend>(
           mgr, config_.model.vits.tokens, meta_data);
-    } else if ((meta_data.is_piper || meta_data.is_coqui) &&
+    } else if ((meta_data.is_piper || meta_data.is_coqui ||
+                meta_data.is_icefall) &&
                !config_.model.vits.data_dir.empty()) {
       frontend_ = std::make_unique<PiperPhonemizeLexicon>(
           mgr, config_.model.vits.tokens, config_.model.vits.data_dir,
@@ -231,7 +232,8 @@ class OfflineTtsVitsImpl : public OfflineTtsImpl {
     if (meta_data.frontend == "characters") {
       frontend_ = std::make_unique<OfflineTtsCharacterFrontend>(
           config_.model.vits.tokens, meta_data);
-    } else if ((meta_data.is_piper || meta_data.is_coqui) &&
+    } else if ((meta_data.is_piper || meta_data.is_coqui ||
+                meta_data.is_icefall) &&
                !config_.model.vits.data_dir.empty()) {
       frontend_ = std::make_unique<PiperPhonemizeLexicon>(
           config_.model.vits.tokens, config_.model.vits.data_dir,

diff --git a/sherpa-onnx/csrc/offline-tts-vits-model-metadata.h b/sherpa-onnx/csrc/offline-tts-vits-model-metadata.h
@@ -20,6 +20,7 @@ struct OfflineTtsVitsModelMetaData {
 
   bool is_piper = false;
   bool is_coqui = false;
+  bool is_icefall = false;
 
   // the following options are for models from coqui-ai/TTS
   int32_t blank_id = 0;

diff --git a/sherpa-onnx/csrc/offline-tts-vits-model.cc b/sherpa-onnx/csrc/offline-tts-vits-model.cc
@@ -110,6 +110,10 @@ class OfflineTtsVitsModel::Impl {
     if (comment.find("coqui") != std::string::npos) {
       meta_data_.is_coqui = true;
     }
+
+    if (comment.find("icefall") != std::string::npos) {
+      meta_data_.is_icefall = true;
+    }
   }
 
   Ort::Value RunVitsPiperOrCoqui(Ort::Value x, int64_t sid, float speed) {

diff --git a/sherpa-onnx/csrc/piper-phonemize-lexicon.cc b/sherpa-onnx/csrc/piper-phonemize-lexicon.cc
@@ -236,7 +236,7 @@ std::vector<std::vector<int64_t>> PiperPhonemizeLexicon::ConvertTextToTokenIds(
 
   std::vector<int64_t> phoneme_ids;
 
-  if (meta_data_.is_piper) {
+  if (meta_data_.is_piper || meta_data_.is_icefall) {
     for (const auto &p : phonemes) {
       phoneme_ids = PiperPhonemesToIds(token2id_, p);
       ans.push_back(std::move(phoneme_ids));

diff --git a/sherpa-onnx/csrc/session.cc b/sherpa-onnx/csrc/session.cc
@@ -105,11 +105,13 @@ static Ort::SessionOptions GetSessionOptionsImpl(int32_t num_threads,
       } else {
         SHERPA_ONNX_LOGE("Use nnapi");
       }
-#else
+#elif defined(__ANDROID_API__)
       SHERPA_ONNX_LOGE(
           "Android NNAPI requires API level >= 27. Current API level %d "
           "Fallback to cpu!",
           (int32_t)__ANDROID_API__);
+#else
+      SHERPA_ONNX_LOGE("NNAPI is for Android only. Fallback to cpu");
 #endif
       break;
     }